PAI-Bench: direct-parent identifiers appear in 48/48 atomic answers and 1/48 self-portraits.

A provider-neutral benchmark separating identity recall from composition, behavioral enactment, resistance, persistence, lineage and role-conditioned updates, with scoring oracles kept outside the target process (arXiv 2609.13637). Two frozen campaigns over sixteen synthetic profiles, thirty-two probes and three independently initialized configurations, 1,536 retained responses. Explicit field cues moved joint presence of three identity identifiers from 0/8 to 7/8 under otherwise identical instructions, and replaying identical factorial responses gave a Claude headline mean 12.5 percentage points below Astra's, which isolates evaluator sensitivity from target behavior.